AI HR assistant for employee questions
The agent nobody puts in the business case and everybody uses. HR fields the same twenty questions from every employee, forever, and almost none of them need a person.
The agent nobody puts in the business case and everybody uses. HR fields the same twenty questions from every employee, forever, and almost none of them need a person.
What this agent is for
Internal assistants have a different risk profile from customer-facing ones and a much clearer return. The population is known and authenticated, the questions are highly repetitive, and the material — policies, handbooks, procedures — already exists in a form the assistant can use.
What makes HR distinctive is sensitivity. The questions people ask an HR assistant are often ones they would rather not ask a colleague: leave entitlement, salary queries, grievance procedures, benefits they are unsure they qualify for. That privacy is a genuine feature — people ask an assistant things they would not raise in person — and it is also why the boundaries have to be unusually well drawn.
The questions it handles
How much leave do I have?
Balance for the person asking, from the system of record, without emailing anyone.
What is the policy on X?
Answered from the handbook, consistently, rather than differently depending on who was asked.
How do I claim expenses?
The actual process in order, with the forms and thresholds that apply.
What are my benefits?
Entitlements for this employee, where the systems allow, rather than a general statement covering everyone.
How do I raise a concern?
The grievance process, explained clearly and without judgement — a question many people will only ask something anonymous.
When is payday?
The kind of question that is trivial, constant, and currently answered by a person.
A typical conversation
- The employee is already knownIdentity comes from your sign-in, so the assistant knows who is asking and can answer about their own entitlements rather than in generalities.
- They ask in plain languageOften about something they are unsure of, sometimes about something they would rather not ask a manager.
- It answers from your policyThe same answer every time, drawn from the current handbook rather than from institutional memory of varying accuracy.
- Personal figures come from the recordLeave balance, entitlement and similar are retrieved for that person, within their own permissions.
- Sensitive topics route to a personGrievances, disputes, anything involving a named colleague, and anything suggesting distress.
- Nothing is quietly logged against themEmployees will not use a system they suspect is reporting their questions to their manager. What is recorded and who can see it should be stated plainly.
How this agent is configured
| Tone | Warm, plain and non-judgemental. People ask HR assistants things they are embarrassed to ask a person, and a brisk tone shuts that down. |
|---|---|
| Knowledge | Handbook, policies, procedures, benefits documentation, and — where connected — the employee's own leave and entitlement records. |
| Escalation triggers | Grievances, anything naming a colleague, disciplinary matters, distress, and any question about an individual's pay beyond published bands. |
| Never states | Anything about another employee, anything that reads as a legal or contractual ruling, and anything not supported by current policy. |
| Privacy posture | What is logged, for how long, and who can see it — stated plainly to employees, because adoption depends on it. |
| Languages | Arabic and English, which matters more internally than externally in most Saudi organisations given mixed workforces. |
Where it hands over to a person
The escalation boundary in HR is narrower and firmer than in most agents. Anything involving a named colleague, any grievance, anything disciplinary, and anything suggesting distress goes to a person immediately and without attempting to help first.
The reason is not capability. The assistant could summarise a grievance procedure perfectly well. The reason is that a person raising a grievance is signalling that they need to be heard by someone, and being answered by an automated system in that moment is experienced as being dismissed — which converts a manageable situation into a formal one.
Pay is the other firm boundary. Published bands and processes are fine. Anything about an individual's specific compensation, or comparisons between people, belongs with HR and nowhere else.
What it should never do
Discuss another employee
Under any framing, including from a manager. Entitlement to information about a colleague is a decision a person makes.
Handle a grievance
A grievance is a request to be heard. Answering it automatically is experienced as dismissal.
Rule on a dispute
Explaining a policy is not the same as applying it to a contested situation.
Discuss individual pay
Published bands and processes only. Anything specific belongs with HR.
Give legal advice
Employment law questions are explained by reference to policy, not answered as legal positions.
Log questions against people
If employees suspect their questions are visible to their manager, they will stop asking, and the deployment fails quietly.
Why a separate agent rather than one that does everything
The instinct is usually to build one assistant that handles everything, on the grounds that customers do not care about your org chart. The instinct is wrong, and for a specific reason: tone, knowledge and risk are different per function, and a single agent forces you to average across all three.
Knowledge is the obvious one. The material a support agent should answer from is not the material a sales agent should answer from, and mixing them means each answers from a corpus containing a large amount of irrelevant text. Retrieval quality falls as the corpus widens, so a general assistant is measurably worse at every individual job than a narrow one.
Tone is less obvious and matters more than people expect. The register appropriate for chasing a payment is not the register appropriate for reassuring someone whose service is down. In Arabic this is sharper still, because formality carries more weight and getting it wrong reads as rudeness rather than as informality.
Risk is the one that decides it. Different functions need different limits — what a finance agent may state about an account is different from what an HR agent may state about an employee, and those limits are far easier to set, test and audit per agent than as conditions inside one large one. Separate agents make the question "what is this allowed to say?" answerable.
How this agent should be measured
| Resolution without escalation | Useful, but never on its own. An agent that makes escalation hard scores well here while the relationship deteriorates — and the deterioration surfaces months after the metric improves. |
|---|---|
| Repeat contact within 48 hours | The number that exposes false resolutions. A conversation marked resolved that returns tomorrow was not resolved; it was deferred. |
| Abandonment mid-conversation | Where people gave up. Rises with latency, with menu depth, and with any answer that reads as evasive rather than unhelpful. |
| Escalations requiring repetition | The share of handovers where the customer had to say something twice. This is the single clearest measure of whether handover works. |
| Answer grounding | What proportion of answers trace to a source document. The number a regulator or an internal auditor eventually asks for. |
| Language accuracy | Whether replies came back in the language of the question. The most common complaint about bilingual assistants, and the easiest to instrument. |
| Satisfaction after escalation | Whether people who reached a person were satisfied. If this is high and containment is low, the deployment is working correctly. |
Mistakes worth avoiding when configuring it
Feeding it marketing copy
The questions people ask are answered in policies, procedures and price lists — not on the brochure page. Answer quality tracks the quality of source material more closely than anything else you control.
Writing the answers before the limits
Deciding what the agent must never attempt takes half an hour and prevents the category of incident that cancels projects. Deciding what it should answer takes far longer and matters less.
Optimising for containment
It is the easiest number to improve for the wrong reasons. Make escalation difficult and containment rises while satisfaction falls.
Translating the persona
An English persona rendered literally into Arabic reads as curt; an Arabic one rendered into English reads as ornate. Configure each language rather than translating one.
Launching broad
Narrow deployments that answer a handful of questions extremely well build trust. Broad ones that answer everything adequately lose it at the first confident mistake.
Leaving it unowned
Agents without a named owner drift — material goes stale, tone wanders, and nobody notices until quality has visibly fallen.
What the person on the other side experiences
It is worth thinking about this agent from the customer's side rather than from the configuration screen, because the two views produce different priorities.
From the configuration screen, success looks like coverage: more questions answered, fewer escalations, a rising containment figure. From the customer's side, success looks like something much simpler — they asked, they got a straight answer, and it did not take long. They do not know or care whether the answer came from an assistant or a person, and they will not credit you for automation. They will, however, notice immediately if the answer was wrong, evasive, or in the wrong language.
This is why the honest failure modes matter so much. "I do not have that information, let me get someone who does" costs you a containment point and buys you trust. A confident wrong answer saves the containment point and costs you the relationship, and you will not find out for weeks.
It is also why language handling is disproportionately important here. A customer who writes in Arabic and is answered in English has been told, without anyone saying it, that they are being served by a system built for someone else. In this market that is not a minor usability issue; it is a statement about who the service was designed for.
Typical impact
Setting it up well
Be explicit with employees about what is recorded. Adoption of an internal assistant depends almost entirely on whether people believe their questions are private. Say what is logged, for how long, and who can see it — in both languages — before launch rather than in response to a rumour.
Start with leave and policy. They are the highest-volume, lowest-risk questions in every organisation, they are answered from material that already exists, and they demonstrate value in the first week.
Do not let it drift into decisions. An HR assistant that explains policy is useful. One that appears to apply policy to a specific situation is making a determination, and employees will treat its answer as authoritative in a dispute. Keep the line clear and state it in the agent's own wording.
Common questions
Only if you are explicit about what is logged and who can see it. That is the single factor that determines adoption of an internal assistant, and it has to be settled before launch.
For the person asking, where the system is connected. Identity comes from your own sign-in, and it answers within that person's permissions.
Straight to a person, without attempting to help first. Someone raising a grievance needs to be heard, and an automated answer in that moment is experienced as being dismissed.
Not unless you deliberately design it that way, and we would advise against it. Suspicion that questions are visible upward stops people asking.
That is often where the value is highest — people who are never in an office, at hours when HR is closed, on a phone, in whichever language they prefer.
It can explain the process and show a balance. Approving is a decision with contractual weight and stays with a person unless you deliberately enable otherwise.
Configure your own agents
One assistant per function, each with its own voice and limits.